Skip to Content Skip to Footer
Applying Machine Learning to the Challenge of Cancer Patient Outreach

Applying Machine Learning to the Challenge of Cancer Patient Outreach

Yixing Chen, Ju-Yeon Lee, Shrihari (Hari) Sridhar, Vikas Mittal, Katharine McCallister and Amit G. Singal

Listen to the authors present their findings (source: March 2020 JM Webinar)

In 2018, over 1.7 million new cases of cancer were diagnosed in the United States and the cost of cancer care surpassed $147 billion. Many of these cases could have been prevented through regular cancer screening tests that open the door for early detection, more cost-effective treatment options, and better recovery prognosis. For example, regular screening reduces mortality rates for lung cancer by 28%, breast cancer by 24%, and liver cancer by 37%. Moreover, cancer screening can reduce the annual treatment cost for a patient by nearly $5,000.

Healthcare institutions rely on marketing interventions—or direct-to-patient outreach—to increase screening completion among at-risk patients. As an example, Johns Hopkins Hospital’s cancer center uses emails, letters, seminars, and community events to encourage screening completion among patients. Yet, according to a recent article in the LA Times, “just 4.2% of patients in the United States who are at high risk for lung cancer get screened for it — a figure seen as alarmingly low by those who work in the area of prevention.” Could it be that the 1.7 million outreach interventions launched in 2015 and $123 million spent on prevention and education efforts go to waste?


A new study in the Journal of Marketing explores the efficacy of patient marketing. Our research team used a multi-period randomized field experiment conducted at University of Texas Southwestern Medical Center in Dallas with at-risk patients for hepatocellular carcinoma (HCC), the most common type of primary liver cancer. Patients were randomly assigned (1:1:1) to three different conditions—usual care, outreach alone, or outreach with patient navigation. Usual care is the baseline condition where physicians offer preventive care recommendations at their discretion during a patient’s usual care visits. Outreach alone and outreach with patient navigation provide two different levels of direct marketing efforts based on outreach mails, outreach calls, and customized motivational education by trained patient navigators. The focal outcome is the patient’s screening completion status within 6 months (Period 1), 6-12 months (Period 2), and 12-18 months (Period 3) of the initial randomization. To incorporate patient heterogeneity, we included patients’ demographics, health status, visit history, health system accessibility, neighborhood socioeconomic status, and prior screening compliance.

We apply causal forests, a state-of-the-art development in machine learning, to discover that the effectiveness of outreach programs varies widely over time and across patients. For example, outreach programs in general are more effective for patients who are female, minority, in better health status, have a more frequent visit history, covered by medical-assistance insurance, reside in closer proximity to clinics, and reside in a more populated neighborhood. Outreach alone is more effective for patients who are younger, commute faster, and reside in a neighborhood with more public insurance coverage. In contrast, outreach with patient navigation is more effective for patients who are older and reside in a higher-income neighborhood.

Incorporating these patient-level differences in their responsiveness to outreach interventions and a well-established scheme of cost-benefit calculation that quantifies health benefits and financial costs associated with outreach interventions, we show that a targeted outreach program (that matches each patient to the optimal outreach type) improves the return on the randomized outreach program (the current state of practice) by 74%-96%, resulting in a gain of $1.6 million to $2 million. Collectively, we help healthcare practitioners in the following ways: 

  • We offer an understanding of how outreach effectiveness varies over time and by patient characteristics that are theoretically relevant and easily accessible to practitioners in their patient database. 
  • We offer a cost-benefit analysis approach for assessing individual-level return on marketing investments in the context of liver cancer screening outreach. 
  • We contribute to the practice of individually and dynamically personalized healthcare marketing by providing a tool that can recommend the most suitable intervention for each patient.

In sum, we combine the well-known marketing principle that all customers are different with advanced machine learning to show that personalized cancer outreach can save both lives and money. 

Read the full article

From: Yixing Chen, Ju-Yeon Lee, Shrihari Sridhar, Vikas Mittal, and Amit G. Singal, “Improving Cancer Outreach Effectiveness Through Targeting and Economic Assessments: Insights from a Randomized Field Experiment,” Journal of Marketing.

Go to the Journal of Marketing

Yixing Chen is Doctoral Student in Marketing, Texas A&M University, Mays School of Business.

Ju-Yeon Lee is Assistant Professor of Marketing, Department of Marketing, Iowa State University.

Shrihari (Hari) Sridhar Associate Professor of Marketing, Center for Executive Development Professor, Texas A&M University, Mays School of Business.

Vikas Mittal is J. Hugh Liedtke Professor of Marketing, Rice University, USA.

Katharine McCallister is Population Science Program Manager, University of Texas Southwestern Medical Center.

Amit G. Singal is Associate Professor of Internal Medicine, Clinical Chief of Hepatology, University of Texas Southwestern Medical Center.